modified for DB functionality and accessible parameters by Mr_KrzYch00.
The AI tries to predict the continuation of the text that has been typed in the Input TEXT window
Generated samples are session locked within browser
774M model (fast) and 1558M (slow, not always available) generate 1 sample in real time
All parameters are adjustable: Temperature, TOP_K and TOP_P
TOP_P is multi-ranged, allowing precise elimination to, for example, shape the output into Human way of writing as close as possible
Both models generate and output data in 20 tokens batches, using 1024-tokens sliding window of input+sample
Option to disable <|endoftext|>, so that the text is continuous
Option to prevent repetitions (new sample+same length part of input+sample to find in 2nd part of processed input+sample -> disallow)
Option to Improve continuation, disallow ending sentence as 20th (< if repeats) token
Option to Provoke continuation, allow ending sentence only as 1st token
Sample generation can be canceled
If 774M GPU Mode OOM's 2 times, the input will be continued by 1558M CPU Mode
Both models are auto-restarted every 50,000 output tokens returned
Mouse-over on each element for the standard OS's tooltip with additional information
This site generates no additional costs . . . (no support is required)
All windows within this site's view can be repositioned freely. Bottom right corner is used to resizes them
Using Nvidia GTX 1660 6GB & AMD Ryzen 1600X
Old projects and stuff here
Please wait . . .
▲ Advanced Parameters ▲
TOP_P List (0 ~ 1) |
0 ~ 0.9
Remove repeated text
▲ Input TEXT ▲
Send to 774M | fast [test mode, seed: 1]
Send to 1558M | slow
▲ Output SAMPLES ▲
Generated samples list (click buttons below to open AI reply)
This is a research site, do not abuse!
The administrator is not responsible for any input user writes as well as the output the algorithm produces.